Sign Up to our social questions and Answers Engine to ask questions, answer people’s questions, and connect with other people.
Login to our social questions & Answers Engine to ask questions answer people’s questions & connect with other people.
Lost your password? Please enter your email address. You will receive a link and will create a new password via email.
Please briefly explain why you feel this question should be reported.
Please briefly explain why you feel this answer should be reported.
Please briefly explain why you feel this user should be reported.
Initially I implemented a Backpropagation network in Matlab and used it on XOR. However, now I am using to the same network using the following input/target combination.
Inputs = [0 0; 0 1; 1000 0; 1 1], Targets = [0; 1000; 1; 0]
And I get output as [1;1;1;1]
The network wasn’t able to learn the network at all. Could anyone please explain why it is so? And what can I do if I to build a network that can learn such small I/O networks?
Any explanation is highly appreciated.
Regards
Max
It’s looks like a scaling problem. In your original XOR problems, the inputs and outputs were all on comparable scales, namely [0,1]. In your revised problem, some inputs appear to be [0,1] and some [0,1000].
The solution is to normalize the inputs to similar scales: [0,1] or [-1,1] are commonly used.
In your case, it may be sufficient to divide the inputs by 1000 to place your range into [0,1]. Don’t forget to denormalize the outputs (i.e. multiply by 1000 in your case) to return to the original scale.